Abstract
With the two goals of providing (a) an introduction to the technology of neural networks and (b) an argument for the admissibility of machine opinion, this Article introduces the technology by first looking at the relatively familiar operation of technology-assisted review (TAR) of documents in a typical case. The Article then outlines the extensive application of neural networks in the real world, which is used later to argue that systems trusted in the field should be trusted in court. The Article then turns to the law of evidence, focusing on the rules governing the admissibility of computer-stored and computer-generated data, including animations and simulations. The theme of those sections is, again, that reliability drives admissibility. This sets the stage for the Article’s central contention, made through four arguments, that the output of neural networks be admissible in court. The Article ends by invoking the need for meaningful cross-examination, setting out the risks—and so the likely targets of that crossexamination—which attend the admission of opinions generated by neural networks.